In this demonstration, we present a unified framework for semantic shot classification in sports videos. Unlike previous approaches, which focus on clustering by aggregating shots with similar low-level features, the proposed scheme makes use of domain knowledge of specific sport to perform a top-down video shot classification. That is, combining with inherent game rules and television field production, for each sport through careful observations we predefine a set of semantic shots which cover 90 to 95 % of sports broadcasting video. Under the supervision of predefined shots set, we map the low-level features to high-level semantic video shot attributes such as dominant object motion (a player), persistent camera panning, and court shape. On the basis of the appropriate fusion of those high-level shot attributes, we classify video shots into several predefined categories, each of which has a clear semantic meaning. The experiments show that, compared to traditional clustering methods and key-frame based analysis, the proposed framework features great capability of semantics mining. Due to remarkable structure constraints and limited sports photography, this framework provides a generic solution for sports video shot classification, which can be adapted to a new sport type without major modification. With correctly classified sports video shots further structural and temporal analysis will be greatly facilitated.

A well-produced video always creates a strong impression on the viewer. However due to the limitations of the camera, the ambient conditions or the skills of the videographer, the quality of captured videos sometimes falls short of one's expectations. On the other hand, we have a vast amount of superbly captured videos available on the web and in digital libraries. In this paper, we propose the novel approach of video analogies that provides a powerful ability to improve the quality of a video by borrowing features from a higher quality video. We want to improve the given target video in order to obtain a higher quality output video. During the matching phase, we find the correspondence between the pair by using feature matching. Then for the target video, we utilize this correspondence to transfer some desired traits of the source video into the target video in order to obtain a new video. Thus the new video will obtain the desired features from the source video while retaining the merits of the target video. The video analogies technique provides an intuitive mechanism for automatic editing of videos. We demonstrate the utility of the analogies method by considering three applications - colorizing videos, reducing video blurs and video rhythm adjustment. We describe each application in detail and provide experimental results to establish the efficacy of the proposed approach.